forked from cmars/conflux
/
peer_gossip.go
303 lines (271 loc) · 8.41 KB
/
peer_gossip.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
/*
conflux - Distributed database synchronization library
Based on the algorithm described in
"Set Reconciliation with Nearly Optimal Communication Complexity",
Yaron Minsky, Ari Trachtenberg, and Richard Zippel, 2004.
Copyright (c) 2012-2015 Casey Marshall <cmars@cmarstech.com>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, version 3.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
package recon
import (
"bufio"
"errors"
"fmt"
"math/rand"
"net"
"time"
log "github.com/sirupsen/logrus"
"gopkg.in/errgo.v1"
cf "github.com/simia-tech/conflux"
)
const GOSSIP = "gossip"
// Gossip with remote servers, acting as a client.
func (p *Peer) Gossip() error {
rand.Seed(time.Now().UnixNano())
timer := time.NewTimer(time.Second * time.Duration(rand.Intn(p.settings.GossipIntervalSecs)))
for {
select {
case <-p.t.Dying():
return nil
case <-timer.C:
if p.readAcquire() {
peer, err := p.choosePartner()
if err != nil {
if errgo.Cause(err) == ErrNoPartners {
p.log(GOSSIP).Debug("no partners to gossip with")
} else {
p.logErr(GOSSIP, err).Error("choosePartner")
}
} else {
err = p.InitiateRecon(peer)
if errgo.Cause(err) == ErrPeerBusy {
p.logErr(GOSSIP, err).Debug()
} else if err != nil {
p.logErr(GOSSIP, err).Errorf("recon with %v failed", peer)
}
}
p.wg.Done()
}
delay := time.Second * time.Duration(rand.Intn(p.settings.GossipIntervalSecs))
p.log(GOSSIP).Infof("waiting %s for next gossip attempt", delay)
timer.Reset(delay)
}
}
}
var ErrNoPartners error = errors.New("no recon partners configured")
var ErrIncompatiblePeer error = errors.New("remote peer configuration is not compatible")
var ErrPeerBusy error = errors.New("peer is busy handling another request")
var ErrReconDone = errors.New("reconciliation done")
func IsGossipBlocked(err error) bool {
switch err {
case ErrNoPartners:
return true
case ErrIncompatiblePeer:
return true
case ErrPeerBusy:
return true
}
return false
}
func (p *Peer) choosePartner() (net.Addr, error) {
partners, err := p.settings.PartnerAddrs()
if err != nil {
return nil, errgo.Mask(err)
}
if len(partners) == 0 {
return nil, errgo.Mask(ErrNoPartners, IsGossipBlocked)
}
return partners[rand.Intn(len(partners))], nil
}
func (p *Peer) InitiateRecon(addr net.Addr) error {
p.log(GOSSIP).Debugf("initiating recon with peer %v", addr)
conn, err := net.DialTimeout(addr.Network(), addr.String(), 30*time.Second)
if err != nil {
return errgo.Mask(err)
}
defer conn.Close()
remoteConfig, err := p.handleConfig(conn, GOSSIP, "")
if err != nil {
return errgo.Mask(err)
}
// Interact with peer
return p.clientRecon(conn, remoteConfig)
}
type msgProgress struct {
elements *cf.ZSet
err error
flush bool
messages []ReconMsg
}
func (mp *msgProgress) String() string {
if mp.err != nil {
return fmt.Sprintf("err=%v", mp.err)
}
return fmt.Sprintf("nelements=%d flush=%v messages=%+v",
mp.elements.Len(), mp.flush, msgTypes(mp.messages))
}
func msgTypes(messages []ReconMsg) []string {
var result []string
for _, msg := range messages {
result = append(result, msg.MsgType().String())
}
return result
}
type msgProgressChan chan *msgProgress
func (p *Peer) clientRecon(conn net.Conn, remoteConfig *Config) error {
w := bufio.NewWriter(conn)
respSet := cf.NewZSet()
defer func() {
p.sendItems(respSet.Items(), conn, remoteConfig)
}()
var pendingMessages []ReconMsg
for step := range p.interactWithServer(conn) {
if step.err != nil {
if step.err == ErrReconDone {
p.log(GOSSIP).Info("reconcilation done")
break
} else {
err := WriteMsg(w, &Error{&textMsg{Text: step.err.Error()}})
if err != nil {
p.logErr(GOSSIP, err).Error()
}
p.logErr(GOSSIP, step.err).Error("step error")
break
}
} else {
pendingMessages = append(pendingMessages, step.messages...)
if step.flush {
for _, msg := range pendingMessages {
err := WriteMsg(w, msg)
if err != nil {
return errgo.Mask(err)
}
}
pendingMessages = nil
err := w.Flush()
if err != nil {
return errgo.Mask(err)
}
}
}
p.log(GOSSIP).Debugf("add step: %v", step)
respSet.AddAll(step.elements)
p.log(GOSSIP).Infof("recover set now %d elements", respSet.Len())
}
return nil
}
func (p *Peer) interactWithServer(conn net.Conn) msgProgressChan {
out := make(msgProgressChan)
go func() {
defer close(out)
var resp *msgProgress
var n int
for (resp == nil || resp.err == nil) && n < maxRecoverSize {
p.setReadDeadline("CLIENT", conn, defaultTimeout)
msg, err := ReadMsg(conn)
if err != nil {
p.logErr(GOSSIP, err).Error("interact: read msg")
out <- &msgProgress{err: err}
return
}
p.logFields(GOSSIP, log.Fields{"msg": msg}).Debug("interact")
switch m := msg.(type) {
case *ReconRqstPoly:
resp = p.handleReconRqstPoly(m)
case *ReconRqstFull:
resp = p.handleReconRqstFull(m)
case *Elements:
p.logFields(GOSSIP, log.Fields{"nelements": m.ZSet.Len()}).Debug()
resp = &msgProgress{elements: m.ZSet}
case *Done:
resp = &msgProgress{err: ErrReconDone}
case *Flush:
resp = &msgProgress{elements: cf.NewZSet(), flush: true}
default:
resp = &msgProgress{err: errgo.Newf("unexpected message: %v", m)}
}
n += resp.elements.Len()
out <- resp
}
}()
return out
}
var ErrReconRqstPolyNotFound = errors.New(
"peer should not receive a request for a non-existant node in ReconRqstPoly")
func (p *Peer) handleReconRqstPoly(rp *ReconRqstPoly) *msgProgress {
remoteSize := rp.Size
points := p.ptree.Points()
remoteSamples := rp.Samples
node, err := p.ptree.Node(rp.Prefix)
if err == ErrNodeNotFound {
return &msgProgress{err: ErrReconRqstPolyNotFound}
}
localSamples := node.SValues()
localSize := node.Size()
remoteSet, localSet, err := p.solve(
remoteSamples, localSamples, remoteSize, localSize, points)
if errgo.Cause(err) == cf.ErrLowMBar {
p.log(GOSSIP).Info("ReconRqstPoly: low MBar")
if node.IsLeaf() || node.Size() < (p.settings.ThreshMult*p.settings.MBar) {
p.logFields(GOSSIP, log.Fields{
"node": node.Key(),
}).Info("sending full elements")
elements, err := node.Elements()
if err != nil {
return &msgProgress{err: errgo.Mask(err)}
}
return &msgProgress{elements: cf.NewZSet(), messages: []ReconMsg{
&FullElements{ZSet: cf.NewZSet(elements...)}}}
} else {
err = errgo.Notef(err, "bs=%v leaf=%v size=%d", node.Key(), node.IsLeaf(), node.Size())
}
}
if err != nil {
p.logErr(GOSSIP, err).Info("ReconRqstPoly: sending SyncFail")
return &msgProgress{elements: cf.NewZSet(), messages: []ReconMsg{&SyncFail{}}}
}
p.logFields(GOSSIP, log.Fields{"localSet": localSet, "remoteSet": remoteSet}).Info("ReconRqstPoly: solved")
return &msgProgress{elements: remoteSet, messages: []ReconMsg{&Elements{ZSet: localSet}}}
}
func (p *Peer) solve(remoteSamples, localSamples []*cf.Zp, remoteSize, localSize int, points []*cf.Zp) (*cf.ZSet, *cf.ZSet, error) {
var values []*cf.Zp
for i, x := range remoteSamples {
values = append(values, cf.Z(x.P).Div(x, localSamples[i]))
}
p.logFields(GOSSIP, log.Fields{
"values": values,
"points": points,
"degDiff": remoteSize - localSize,
}).Debug("reconcile")
return cf.Reconcile(values, points, remoteSize-localSize)
}
func (p *Peer) handleReconRqstFull(rf *ReconRqstFull) *msgProgress {
var localset *cf.ZSet
node, err := p.ptree.Node(rf.Prefix)
if err == ErrNodeNotFound {
localset = cf.NewZSet()
} else if err != nil {
return &msgProgress{err: err}
} else {
elements, err := node.Elements()
if err != nil {
return &msgProgress{err: err}
}
localset = cf.NewZSet(elements...)
}
localNeeds := cf.ZSetDiff(rf.Elements, localset)
remoteNeeds := cf.ZSetDiff(localset, rf.Elements)
p.logFields(GOSSIP, log.Fields{
"localNeeds": localNeeds.Len(),
"remoteNeeds": remoteNeeds.Len(),
}).Info("ReconRqstFull")
return &msgProgress{elements: localNeeds, messages: []ReconMsg{&Elements{ZSet: remoteNeeds}}}
}